python sklearn 正向激励决策树 特征重要性

python sklearn 正向激励决策树 特征重要性

"""
    决策树模型 正向激励 
    特征重要性 特征对结果的影响力 
"""
import sklearn.datasets as sd
import sklearn.utils as su
import sklearn.tree as st
import sklearn.metrics as sm
import sklearn.ensemble as se
import matplotlib.pyplot as mp
import numpy as np

boston = sd.load_boston()
print(boston.data.shape)
print(boston.target.shape)
print(boston.feature_names)
print(boston.data[0])
print(boston.target[0])
names = boston.feature_names

# 1.准备数据
x, y = su.shuffle(boston.data, boston.target, random_state=7)
train_size = int(len(x) * 0.8)
train_x, train_y, test_x, test_y = \
    x[:train_size], y[:train_size], x[train_size:], y[train_size:]

# 2.决策树模型
model = st.DecisionTreeRegressor(max_depth=4)
model.fit(train_x, train_y)
prd_test_y = model.predict(test_x)

# 3.决策树模型 --评估
print("R2得分:", sm.r2_score(test_y, prd_test_y))
print("平均绝对值误差:", sm.mean_absolute_error(test_y, prd_test_y))

dt_fi = model.feature_importances_  # 特征重要性
print("----特征重要性:\n", dt_fi)

# 4.决策树增加 正向激励
print("--" * 30, "增加正向激励")
model = st.DecisionTreeRegressor(max_depth=4)
model = se.AdaBoostRegressor(model, n_estimators=400, random_state=7)  # 正向激励
model.fit(train_x, train_y)
prd_test_y = model.predict(test_x)

# 5.决策树增加 正向激励 --评估
print("R2得分:", sm.r2_score(test_y, prd_test_y))
print("平均绝对值误差:", sm.mean_absolute_error(test_y, prd_test_y))

ada_fi = model.feature_importances_    # 特征重要性
print("----Adaboost 特征重要性:\n", dt_fi)

# 6.绘图
mp.figure("Feature importance", facecolor="lightgray")

# 7.决策树模型 特征重要性图
mp.subplot(211)
mp.title("DT Feature importance", fontsize=16)
mp.ylabel("Feature importance", fontsize=14)
x = np.arange(dt_fi.size)

sorted_index = dt_fi.argsort()[::-1]  # 对特征重要性进行排序
dt_fi = dt_fi[sorted_index]
print(dt_fi)
mp.grid(linestyle=":", axis="y")
mp.bar(x, dt_fi, 0.8, color="dodgerblue", label="DT Feature importance")
mp.xticks(x, names[sorted_index])
mp.legend()
mp.tight_layout()

# 8.决策树模型+正向激励 特征重要性图
mp.subplot(212)
mp.title("Ada Feature importance", fontsize=16)
mp.ylabel("Feature importance", fontsize=14)
x = np.arange(ada_fi.size)

sorted_index = ada_fi.argsort()[::-1] # 对FI进行排序
ada_fi = ada_fi[sorted_index]

mp.bar(x, ada_fi, 0.8, color="orangered", label="Ada Feature importance")
mp.grid(linestyle=":", axis="y")
mp.xticks(x, names[sorted_index])
mp.legend()
mp.tight_layout()

mp.show()